The functional sequence meets the lexicalization problem
2.2 Polysemy in the lexicalization of the functional sequence
2.2.1 Polysemy of phrasal modifiers
Polysemy is widely attested with both adverbs and adjectives. Cinque’s (1999)Adverbs and Func-tional Heads discusses the syntax of adverbs. On the basis of ordering restrictions between func-tional heads on the one hand and adverbs on the other, Cinque sets up a very fine-grained funcfunc-tional sequence betweenv and C. He shows that the hierarchy of adverbs and the independently estab-lished hierarchy of heads show remarkable parallelisms. He accounts for both hierarchies by a single functional sequence, where the adverbs sit in the specifiers of the semantically correspond-ing functional heads.1 (1) shows the order of functional heads and where applicable, the English
1The claim that adverbs are specifiers has not remained uncontested, see esp. Ernst (2002) for a theory of adjunction.
13
adverbs hosted in their specifiers.2
(1) Moodspeech act frankly, honestly, sincerely
Moodevaluative fortunately, luckily, oddly, regrettably Moodevidential allegedly, reportedly, obviously, evidently Modepistemic probably, likely, supposedly, presumably
T(Past) once
T(Future) then
Moodirrealis perhaps
Modaleth necess (not) necessarily Modaleth possib possibly
Asphabitual usually, generally, regularly, customarily
Aspdelayed finally
Asppredispositional
Asprepetitive(I) again
Aspfrequentative(I) often, repeatedly, X times, twice, frequently Modvolition intentionally
Aspcelerative(I) quickly, rapidly T(Anterior) already
Aspterminative no longer Aspcontinuative still
Aspperfect always
Aspretrospective just, recently, lately Aspproximative soon, immediately
Aspdurative briefly
Aspprogressive characteristically
Aspprospective almost, immediately, nearly, imminently Aspinceptive
Modobligation
Modability
Aspfrustrative/success
Modpermission
Aspconative
AspSg.completive(I) completely AspPl.completive
Voice well, manner adverbs
Aspcelerative(II) quickly, rapidly, fast, early Aspinceptive(II)
Asprepetitive(II) again
Aspfrequentative(II) often, repeatedly, X times, twice, frequently AspSg.completive(II) completely
The adverbs in bold can appear in two different places in the hierarchy of adverbs, which gives rise to the impression that their position is not fixed. Again, often, quickly, rapidly and completely thus pose a potential problem to the idea of a universal, rigid functional hierarchy. So do many more adverbs that can appear in two (or more) positions in the clause, such ascleverly, stupidly, tactfully, agressively, rudely, graciously and so on. The standard example to illustrate the problem involvescleverly. It is well-known that the different positions correspond to different interpretations.3
(2) a. John has cleverly answered their questions.
b. John cleverly has answered their questions.
2(1) combines the hierarchy in Cinque (1999, p. 106., ex. 92) with the refinements presented in Cinque (2006f).
3The polysemy of these adverbs has an extensive literature. I refer the reader to McConnell-Ginet (1982);
Geuder (2000); Ernst (2002); Wyner (2009) and Pi˜n´on (2010) for examples and discussion. What is important for our purposes is that in a cartographic view, all these adverbs are capable of occurring in different functional projections.
c. John has answered their questions cleverly.
Cinque (1999, ch. 1., ex. 83.)
Cinque proposes that the functional hierarchy is rigid, after all, and that the relevant adverbs appear in multiple places because they can be base-generated in more than one functional projec-tion. Polysemous adverbs of thecleverly-type, for instance, can be generated in the specifier of the Voice head, in which case they receive a manner interpretation; or in the specifier of a deon-tic Modality head (depending on the adverb, Modvolition, Modobligation or Modability/permission), where they receive a subject-oriented interpretation. Potential support for this comes from data like (3), with both positions filled.
(3) John has cleverly answered the questions cleverly/foolishly.
In a similar fashion,quickly can be the specifier of both a higher Aspcelerative(I) head, where it modifies the event (X is quick in . . . ), and a lower Aspcelerative(II) head, where it modifies the process (X does Y in a quick way). Thus the way polysemous adverbs are interpreted depends on the functional projection they are a specifier of.
Cinque (1999, 2006c) are very clear that these cases involve one and the same adverb base-generated in different positions, rather than homophonous adverbs. Given a systematic polysemy for a class of adverbs, he strictly excludes an ambiguity or homonymy approach. Cinque suggests that these adverbs have a core meaning, which makes them compatible with more than one FP;
and the meaning of the adverb and the meaning of the functional head combine and yield the interpretation together.
The type of highly elaborated functional sequence advocated in Cinque (1999) has been adapted for the analysis of adjectives in Scott (2002). Scott proposes that like adverbs, adjectives also sit in specifiers of functional projections, and proposes the hierarchy in (4) to account for the cross-linguistic patterns of adjective ordering.
(4) ordinal > cardinal > Adjsubjective comment> ?Adjevidential> Adjsize> Adjlength> Adjheight
> Adjspeed > ?Adjdepth > Adjwidth > Adjweight> Adjtemperature > ?Adjwetness> Adjage >
Adjshape> Adjcolor > Adjnationality/origin> Adjmaterial > compound element
Just like certain adverbs seem to occur in more than one position in the hierarchy, so do some adjectives. Scott comments on this in the following way:
Adjectives with the same orthography but which can occur in different positions must be able to be specifiers of more than one (i.e., different) FP.
Scott (2002, p. 105.) The adjectives that can appear in two positions in the adjective hierarchy fall into two classes.
In the first class we find truly ambiguous, homophonous adjectives likecool ‘not hot’ vs. ‘great’
and green ‘the color green’ vs. ‘inexperienced’. The higher and the lower occurrence of these adjectives are entirely unrelated (a young green Martian is green in color, while a green young Martian is inexperienced). For these adjectives, the lexicon contains two different, homophonous items, which are consistently merged in different projections (e.g. SubjectcommentP forcool ‘great’
and TemperatureP forcool ‘not hot’).
In the second class we find adjectives that have a core meaning compatible with more than one adjective-related FP. The adjective old is a case in point. Its core meaning is such that it can be merged in the specifier of an age-related FP (as in an old man) or in the specifier of a temporal-related FP (as in my old (=former) boss).4 Just like with adverbs, the interpretation that these adjectives eventually get depends on what sort of grammatical-semantical information the functional projection adds to their core meaning.5
4Scott tentatively suggests that a similar analysis is also possible for ancient, bulky, ponderous and perhaps tiny, which can be merged either in some age- or size-related and a SubjectcommentP (though he also outlines a possibility whereby these are always merged in SubjectcommentP).
5While Scott explicitly mentions a parallel with the cleverly-type adverbs in the discussion of thegreen-type adjectives, the parallel seems to hold rather with theold-type adjectives, as in both cases the two readings make use of the same lexical item.
2.2.2 Polysemy of heads
In the verbal f-seq, modal verbs, restructuring verbs and light verbs are all representatives of lexical items that occur in various positions with various meanings in the functional sequence. Modalities come in different types: deontic modality is related to will or obligation, epistemic modality is re-lated to knowledge, and alethic modality is rere-lated to necessary and possible truths. Epistemic and deontic modality are widely assumed to be represented by projections in the functional sequence.
It is well known that some languages allow a sequence of two modals (German, Catalan, Spanish, various Scots and American English dialects as well as the Scandinavian languages); and that co-occurring modals have a rigid order: epistemics must precede other modals (c.f. Vikner, 1988;
Brown, 1991; Thr´ainsson and Vikner, 1995; Roussou, 1999, among others). This strict order has lead to the claim that the functional projection for epistemics is higher in f-seq than the projection(s) for deontics (Picallo, 1990; Cinque, 1999, among others).6 Cinque (1999) and Cinque (2006f) argue that alethic modality is also represented in the functional sequence by a dedicated projection, and the order is Modepistemic > Modalethic > Moddeontic.
Modal verbs are often polysemous, and can be used to express two or three different kinds modalities. (5) and (6) from English are exemplar. In the cartographic approach this means that modals can be merged at different points of f-seq.
(5) Kate must be in her office.
deontic: ‘Kate has an obligation to be in her office.’
epistemic: ‘Based on knowledge about the world, the speaker is entirely confident that Kate is in her office.’
(6) John might go home.
epistemic: ‘Based on what the speaker knows, it is possibly true that John will go home.’
alethic: ‘It is possible that John goes home.’ (pure possibility, not related to the speaker’s confidence in the utterance)
deontic: ‘John is allowed to go home.’
The phenomenon of‘restructuring’ first came under discussion in Rizzi (1978), and has gained a lot of attention in the subsequent literature. As pointed out by Cinque, the semantic content of all restructuring verbs is such that it makes them able to represent (i.e. lexicalize) some functional head in (6). The existence of this systematic correspondence led to the claim that restructuring verbs can appear in two distinct positions in the clause: they can be merged either as a main verb in VP (and take a clausal complement) or as a functional verb in the head of the semantically corresponding functional projection (Cinque, 2001, reprinted as Cinque, 2006f; Cinque, 2003, reprinted as Cinque, 2006b; and Cardinaletti and Schlonsky, 2004). The former case results in a biclausal configuration with no transparency effects. In the latter case the structure is monoclausal and transparency effects obtain.
Cinque (2004), reprinted as Cinque (2006e) rejects this proposal and argues that restructuring verbs are always merged as functional heads. But this still does not mean that restructuring verbs are always merged in one and the same position. Restructuring verbs appear to be rigidly ordered, as is expected if they are in the heads of functional projections in (6). Some exceptions exist, however, where one and the same verb can either precede or follow other verbs. This is reminiscent of what we have seen for adverbs, and Cinque argues that it should be treated in the same manner, too. That is, the relevant restructuring verbs can merge in more than one functional projection. The Italian verb cominciare, for instance, can lexicalize both a higher and a lower Inceptive head. This gives rise to the apparently free word order with the heads between the two Inceptive projections.7
Thus there are two kinds of claims about the polysemy of restructuring verbs: some analyses posit a polysemy between a main verb use and a functional verb use; while others claim that
6See, however, Cormack and Smith (2002) for a criticism of this approach and an alternative analysis, and Barbiers (2002) for an overview of various approaches to epistemic and deontic modality.
7See also Fukuda (2008) for puts forward an analysis of English aspectual verbs in the same spirit. He proposes that these verbs can be merged either in the head of a Low AsP phase (belowv), taking a gerundive complement, or in a High Asp head (abovev), taking an infinitival complement. Aspectual verbs likebegin, start, continueand ceasecan take either infinitive or gerundive complements, thus – Fukuda suggests – they can be merged in either in the higher or the lower Asp head.
polysemy stems from merger in different functional projections.
So-called light verbs are another eminent example of lexical items occurring in more than one position in the functional sequence. Light verbs mostly contribute aspectual or aktionsart information to the clause and form a complex predicate with the main verb in the sentence.8 They occur in a dedicated slot in the verbal sequence that is different from the position of both main verbs and auxiliaries (Butt and Lahiri, 2002).
Butt and Lahiri (2002) propose a cross-linguistic generalization about light verbs: in every language that makes use of them, light verbs have main verb uses as well. This observation became known as Butt’s generalization.
(7) Butt’s generalization
A light verb is always form-identical with a main verb in the language.
The following Malayalam (Dravidian) examples testify to the dual nature of light verbs.
(8) a. avan he
kada-(y)il store-loc
pooyi went
‘He went to the store.’ (Rosmin Mathew, p.c.) b. kuppi
bottlepot.t.I break-cp
pooyi GO-pst
‘(The) bottle broke.’ (Abbi and Gopalakrishnan, 1991, p. 162.) Malayalam
On a widespread conception of light verbs, they have undergone grammaticalization and/or semantic bleaching, and they are derived from the corresponding main verb. However, as Butt and Lahiri (2002); Butt (2003); Butt and Geuder (2003) and Butt (2010) point out, this view cannot account for the observed form-identity. Auxiliaries that have undoubtedly undergone grammati-calization split off from the main verb and tend to undergo a grammatigrammati-calization cline, developing different forms and functions from their source verb. Light verbs never undergo this process, they remain form-identical with the main verb. When the form of the main verb changes, the form of the light verb also undergoes the same change (and vice versa).
This leads Butt (2003); Butt and Lahiri (2002) and Butt (2010) to conclude that the main verb and the light verb make use of the same lexical entry. See also Ramchand (2008a) and Ramchand (2008b, ch. 5.6.) for a syntactic analysis that posits one unified lexical entry for the main verb and the light verb use. If this is on the right track, then light verbs are an exemplar of a large class of verbs that can appear in two positions in the functional sequence.
Finally, polysemy of heads is also attested in the nominal f-seq. Many languages make use of classifiers as categorizing devices in the DP. Classifiers come in various types: there are numeral, noun, genitive, verbal, and locative or deictic classifiers (Aikhenvald, 2000). Example (9) illustrates noun classifiers; these typically sort nouns into categories like man, woman, animal, bird, etc.9 (9) mayi
vegetable-abs jimirr yam-abs
bala-al person-erg
yaburu-Ngu girl-erg
julaal dig-past
‘The person girl dug up the vegetable yam.’ (Dixon, 1982, pg. 185) Yidiny
In some languages all noun classifiers systematically lead doubles lives as noun classifiers and full nouns. Examples include Minangkabau from the Austronesian language family, the Amazo-nian Dˆaw (Aikhenvald, 2000), and a number of Australian languages such as Mparntwe Arrernte (Wilkins, 2000), Yidiny (Dixon, 1982), Yir-Yoront (Alpher, 1991, cited in Wilkins, 2000), Kugu
8Depending on the language, light verbs can form complex predicates with verbs, nouns, adjectives and adpo-sitions. Here I will restrict my attention to light verb – verb complex predicates, as only these are relevant to the discussion.
9In the languages that have them, noun classifiers occur in the nominal phrase independently of any other constituents either inside or outside of the DP; that is, noun classifiers are classifiers that are not in need of any licensor element. This contrasts with numeral classifiers, for instance, the occurrence of which is typically licensed by numerals or quantifiers. Noun classifiers scope over the noun phrase, they do not trigger agreement and their choice is determined by semantics/lexical selection. The meaning relation between the classifier and the noun is often generic-specific, which is why in Australianist linguistics they are referred to as generic classifiers or generics (Aikhenvald, 2000).
Nganhcara (Johnson, 1998, cited in Wilkins, 2000) as well as Limilngan and Wardaman (Sands, 1995).
Wilkins (2000) notes that in these languages, the classifier meaning and the corresponding
‘heavy’ noun meaning are generally treated as separate. He points out that this approach leads to a massive and regular polysemy, which seems to miss an important generalization. Wilkins proposes that the noun classifier use and the noun use feature the same lexical item.10 This analysis is more satisfactory than the generally pursued homonymy approach. If there is indeed a single lexical entry behind the noun classifier use and the noun use, as Wilkins suggests, then this entry can be merged in two different positions in the functional sequence of the above mentioned languages: in the noun position and in the noun classifier position.11
2.2.3 Interim summary and outlook
In the preceding sections, we have seen that polysemy is ubiquitous in the literature, and that compatibility with multiple positions is a frequently used tool of cartographic theory to attack this problem. Explicit discussions of polysemy always invoke the idea that the lexical entry has a rather general core meaning, which makes it semantically compatible with more than one functional projection.
I do not contest the validity of the the ‘general core meaning + merge in various positions’
approach. I believe that it is actually correct for a large set of cases, especially the phrasal modifiers (adverbs and adjectives). But the cases of head-polysemy I will look at will receive a better analysis by a Nanosyntactic approach to lexicalization. Nanosyntax is a lexicalization algorithm that operates with non-terminal spellout and restricts systematic polysemy to subset-superset relations between both meanings and positions of a lexical item. In Chapters 3 through 5, I will show that this approach brings empirical payoffs when applied to both new and old data.
Before turning to the detailed exposition of lexicalization in Nanosyntax in Section 2.4, I will first give a short overview of the various works that operate with non-terminal spellout.